Minimum wage: the price of distrust

Transcription

1 Minimum wage: the price of distrust Yann Algan Université Marne la Vallée, IZA, PSE 1 Pierre Cahuc 2 UniversitéParis1,CREST-INSEE,IZA,CEPR (Incomplete and Preliminary version) June 23, yann.algan@ens.fr. 2 Corresponding author: CREST-INSEE, Timbre J 360, 15, Boulevard Gabriel-Peri, 92245, Malakoff, France. cahuc@ensae.fr.

2 Abstract Our paper documents that minimum wage is higher and union density is lower in countries where individuals distrust each other. We first provide theoretical foundations for these correlations by arguing that minimum wage regulation is used when attitudes towards social cooperation are too weak to sustain powerful labor unions that may improve efficiency through collective bargaining. From this point of view, minimum wage is the price of distrust. Then, we develop an original empirical approach to show that cross-country differences in attitudes cause, at least in the short run, differences in union membership, minimum wage and labor market performance in OECD countries over the period Keywords: Minimum wage, labor union, trust. JEL codes: J52, J53, J80, Z13.

3 % Union density Swd Dk Ger Fin Nw Ire Ita Aus Aut Uk Cd Czr Usa Jp Nth Hg Mx Bg Fra Sp R²=0.52 Pt Gre Pol Minimum wage enforcement Index Figure 1: Union density and the degree of enforcement of the minimum wage. Period Source: OECD and ILO (see appendix). 1 Introduction Why union density and minimum wage legislation are so different across OCDE countries? Why is union density low in countries in which the degree of enforcement of the minimum wage is high as shown by Figure 1? 1 Although an overwhelming literature has stressed the key role played by these wage-setting institutions on labor market performance, little effort has been devoted so far to the documentation and the explanation of the stylized facts displayed in Figure 1. Our paper aims at filling this gap. Regarding minimum wage legislation first, the political economic literature has so far mainly stressed the political power of incumbent employees to document the observed negative correlation between right-wing governments and the growth rate of the minimum wage. 2 But this prediction is of little help for understanding the cross-country heterogeneity of minimum wage 1 The indicator for minimum wage enforcement is a composite index measuring the extent to which the legislation is binding. It includes the existence of a legal statutory minimum wage, the dispersion of wage floors across the country, and the provision of sub-minimum wages. See the appendix A.2 for a more detailed description. 2 See Sobel (1999), Saint-Paul (2001), Bacache-Beauvalet and Lehmann (2005). 1

4 legislation because the political influence of incumbent employees has been so far hardly comparable in an international perspective. Moreover most of this analysis is only based on the level of the minimum wage and leaves unexplained why some countries rely on bargained wage floor between unions instead of implementing statutory minimum wage. 3 The explanation of the observed cross country variation in union membership raises the same concerns. In particular, it is difficult to understand why union density has declined by at least ten percentage points since 1970 in some countries like France, while it has remained constant or even increased in other countries such as Nordic ones (see Blanchflower, 2006). The empirical literature has so far stressed the influence of large trends in unemployment rates, demography, product market competition or in the composition of employment which shifted from highly unionized to traditionally non-union sectors and workers (Boeri et al., 2001, Pencavel, 2005, Blanchflower, 2006). But these variables can be thought as much as consequences as causes of the evolution of union membership. Moreover, the literature is mute regarding the observed cross-country trade-off between union membership and minimum wage legislation. Our paper sheds light on this issue: we argue that the negative international correlation between minimum wage legislation and union density is rooted in the cross-country heterogeneity of attitudes towards social cooperation. Our benchmark explanation is straightforward: as stressed by Akerlof (1980) and many others, 4 labor unions can counteract the potential monopsony power of employers more easily in societies where trust and civic cooperation are strong enough to insure involvement in collective action. Conversely, the political demand for a statutory minimum wage is expected to be higher when attitudes towards social cooperation 5 do not allow workers to sustain powerful trade unions. At first sight, empirical evidence suggests that this explanation is relevant. Figure 2 shows that minimum wage legislation is less stringent in countries in which individuals reply more frequently that most people can be trusted. Figure 3 also shows that this attitude is positively correlated with union density. The paper provides an explanation to these correlations and explores potential causality going from attitudes towards social cooperation to labor market institutions by proceeding in two steps. The first step analyzes the theoretical channel through which attitudes towards social cooperation on one hand and wage setting institutions on the other hand could interact. For that 3 Epstein and Nitzan (1999) provide some theoretical ideas on this issue. 4 In particular, Booth (1985), Booth and Chatterji (1993), Naylor (1989, 1990), Naylor and Cripps (1993), Corneo (1995, 1997). 5 Attitudes towards social cooperation are closely related to the concept of social capital put forward by Putnam (1993, 2000) and Coleman (1990) among many others. For instance, Coleman (1990, p 300) argues that authority relations, relations of trusts and consensual allocations of rights which establish norms can be viewed as resources that help individuals to adopt cooperative behavior. 2

5 Minimum wage enforcement Index PtGre Mx Fra Hg Czr Pol Aut Ger Bg Sp Ita Ire Nth Jp Usa Uk Aus Cd Fin Nw Dk Swd R²= % Trusting people Figure 2: Trust and minimum wage enforcement Source: World Value Survey and ILO. 3

6 Swd R²=0.66 % Union density Pt Hg Gre Czr Mx Fra Pol Aut Ger Bg Ire Ita Sp Nth Jp Usa Uk Aus Cd Fin Dk Nw % Trusting people Figure 3: Trust and union density Source: World Value Survey and OECD. 4

7 purpose we lay down a simple political economy model in which a government which can implement a minimum wage is elected and where trade union membership is endogenous. Employers are assumed to have some monopsony power so that both the minimum wage and unionization can be used as a means to counteract the power of the employers on the labor market. Yet union membership is assumed to be more efficient than the minimum wage in lines with the mechanisms put forward by Freeman and Medoff (1984). The efficiency of a union lies in the fact that it favors the choice of a strategy of voice instead of exit, by transmitting complaints, grievances, and demands, with the aim of correcting and improving labor relationships. 6 In this context, minimum wage entails social costs by reducing the incentives to join trade unions and then reduces the possibility to generate gains through collective bargaining. Therefore, it turns out that elections lead to higher minimum wage and lower union density when attitudes towards social cooperations are weak. The theoretical model also sheds some light on the dynamic interactions between social attitudes, minimum wage and union density providing new insights on the dynamics of social attitudes. It shows that if social attitudes act as constraints on the choice of wage-setting institutions in the short run, the former ones are conversely partly shaped by wage-setting institutions in the long-run. For instance, a contemporaneous increase in the statutory minimum wage would reduce incentives to join labor unions and weaken attitudes towards social cooperation in the long run. 7 The second part of the paper aims at testing the hypothesis and the predictions of the theoretical model. We stress, by using international individual surveys over the period , that individuals who tend to distrust the fairness of others are more likely to support stringent minimum wage legislations and are conversely less likely to be unionized. These individual attitudes are strongly correlated with country fixed effects, even after controlling for individual socioeconomic characteristics, suggesting that the correlation between social cooperation and preferences over wage-setting institutions is largely driven by national specific feature. Next, we examine the extent to which attitudes towards social cooperation affect the current wage-setting environment and labor market performances of OECD countries. By analyzing 22 OECD countries over the period , it first turns out that social attitudes are strongly positively correlated with union density and negatively correlated with the stringency of minimum wage legislations. Obviously, the correlation between social attitudes and the design of labor market institutions does not mean that the causal relation goes from social attitudes to 6 See the survey by Addison and Belfield (2004). 7 See Benabou and Tirole (2006) for a complementary analysis on the interactions between incentives and prosocial behavior. 5

8 labor market institutions. We thus push further the analysis by providing some evidence of such a causal relationship. We show that people who face the same economic environment, by living inthesamecountry,namelytheus,haveattitudesthatarecorrelatedwiththecountryof origin of their ancestors. 8 This correlation allows us to identify the attitudes that are inherited independently of the economic environment. In this context, the correlation between current inherited attitudes of people living in the US and current labor market institutions in the country of origin of their ancestors can be interpreted, with some further restrictions, as a causal relation going from current inherited social attitudes to labor market institutions within each country. 9 This approach allows us to provide evidence that cross-country differences in union membership and minimum wage are shaped, at least in the short run, by differences in attitudes towards social cooperation that change slowly over time. We next extend this identification strategy to show that the cross-country heterogeneity in unemployment rates is significantly influenced by the heterogeneity in inherited national attitudes towards social cooperation. More broadly, this paper suggests that indicators based on individual attitudes towards social cooperation might provide a better understanding of labor market outcomes than the quantitative indicators of labor market institutions currently used by the literature (Nickell et al., 2005). As stressed recently by Blanchard (2005), aggregate indicators are doing a poor job in explaining the evolution of the employment patterns in OECD countries. By contrast, observed individual attitudes towards social cooperation provide a more comprehensive picture of the ability of a country to reach good outcomes in terms of employment or unemployment (see Blanchard and Philippon, 2004, for a similar analysis applied to strikes). Moreover, the current literature explaining employment patterns by labor market institutions is fraught with a clear endogeneity bias since these institutions have also changed in reaction to the evolution of employment patterns. In contrast, our strategy helps us to get rid of this endogeneity bias by identifying the cultural traits in social cooperation which are not directly affected by contemporaneous institutions and economic environment. Eventually, this line of research offers new perspective for understanding the possibility of labor market reforms. It suggests that in- 8 The influence of the country of origin on Americans attitudes towards social cooperation was first stressed in the political science litterature by Rice and Feldman (1997). 9 To that regard our approach shows that social attitudes are not overdetermined by current institutions. As illustrated by our theoretical model, dynamic interactions between social attitudes and institutions imply that countries can experience very different histories. Some countries can start with good initial attitudes towards social cooperation that depreciate over time, leading ultimately to a decline in union membership compensated by more stringent minimum wage regulations. Conversely some countries can start from bad attitudes towards social cooperation but then converge to a situation with good social attitudes, high union density and no legal minimum wage. This result casts some doubt on current estimation strategies which use initial historical institutions as an instrument for current social attitudes (Tabellini, 2005) or contemporaneous institutions (Acemoglu et al., 2006). 6

9 stitutions often blamed to induce poor labor market performance in Continental European and Mediterranean countries are much more the symptoms of the lack of prosocial attitudes rather than the cause of unemployment. This finding calls for a new research agenda on policies which help to forster prosocial behavior. (It suggests that the lack of social cooperation in Continental European and Mediterranean countries is a constraint which prevents the implementation of institutions that improve labor market performance.) 2 Theory We consider a discrete time economy with infinite horizon in which non-overlapping generations of risk neutral workers live one period and offer one unit of labor. The measure of workers is normalized to one. There are two non storable goods: a numeraire good and labor. Within each period, the sequence of decisions is the following: 1. Individuals vote to elect a government that offerstosetaminimumwage w The government sets the minimum wage. 3. Once the minimum wage is set, workers can decide to join trade unions. 4. Wages are set by employers for non unionized workers and by wage negotiation for unionized workers. The model is solved backward. We thus start by describing the outcome of wage setting and the decisions to unionize before moving to the choice of the minimum wage Let us first start with the wage setting process between employers and employees (step 4). Employers benefit from monopsony power that allows them to make take-it-or-leave offers to non unionized workers. The indirect utility of non unionized workers amounts to their wage if they work and to zero otherwise. In this context, employers offer them the minimum wage w 0, set by the government. Workers who decide to join trade unions bear some costs denoted by c 0. These subjective costs of collective action and adhesion to collective organization are assumed to be heterogeneous across workers. The cumulative distribution function of c is denoted by F. Following an idea formulated by Akerlof (1980), and further developed by Booth (1985), Booth and Chatterji (1993), Naylor (1989, 1990), Naylor and Cripps (1993), Corneo (1995, 1997), we assume that union members enjoy some social benefits from complying with a social custom that invokes 7

10 workers to express mutual solidarity by joining collective actions. In order to account for such effects, it is assumed that the non-wage gains from joining a union are influenced by social norms. The benefits from becoming an union member increase with the strength of the attitudes towards social cooperation, denoted by S 0. In this framework, the indirect utility of type-c individuals who get a wage w amounts to ½ w c + S w if unionized otherwise. The wage set by trade unions is a share β [0, 1] of the individual production 10 of unionized workers, denoted by y(s), y 0 (S) 0, y(0) = ȳ, where ȳ denotes the individual production of non unionized workers. In other words, the productivity of unionized workers is assumed to be an increasing function of the social benefits derived from unionization, with a lower bound ȳ equal to the productivity of the non unionized workers. This assumption captures the idea that bargaining with labor unions is more efficient when individuals exhibit more inclination for social cooperation (Freeman and Medof, 1984, Luchak, 2003, Addison and Belfield, 2004). β is a measure of the bargaining power of trade unions, which is considered as an exogenous variable. Depending on the wage setting outcome, workers decide to join unions (step 3) if and only if the utility derived from union membership, equals to w c + S, is larger than the utility obtained without union membership, equal to the minimum wage w. Therefore, union density, denoted by D, amounts to D = F ( c), c = βy(s) w + S. (1) It turns out that union density increases with β, the bargaining power of trade unions, and with S, the strength of attitudes towards social cooperation. In contrast, minimum wage rises decrease union membership. Let now turn to the minimum wage set by the government (step 2). The election process is represented by the probabilistic voting model which implies, under some assumptions assumed to be fulfilled, that the elected government maximizes the sum of the utility of the voters. 11 Accordingly, the government chooses the minimum wage w 0 that maximizes W = Z c 0 [y(s) c + S] df (s)+ Z + c ȳdf(s). (2) 10 Production is interpreted in a wide sense, including production net of utility costs associated with waged work. 11 This outcome can be derived from the simple case in which each group of individuals of type-c is heterogeneous with respect to ideological biases towards the two candidates. Then, following Persson and Tabellini (2000) it turns out that the outcome of the elections maximizes the utilitarian criterion if the ideological bias is represented by an additive term in the utility function and is distributed with a uniform distribution that is the same for all type-c individuals. 8

11 The optimal minimum wage satisfies the first-order condition 12 which can be written as w =max{βy(s)+[ȳ y(s)], 0}. (3) This expression of the minimum wage together with the equilibrium level of trade union density (1) yields the following properties for the statutory minimum wage: Result 1: The optimal minimum wage decreases when attitudes towards social cooperation are stronger. This result simply obtains by deriving the expression of w given equation (3) which implies, when w >0, that d w ds = (1 β)y0 (S) 0. Result 1 can be understood as follows. For any type-c individual, the private gains from joining a trade union correspond to the difference between the indirect utility derived from unionization, βy(s) c + S, and the indirect utility derived without unionization, w. The net social gains of the involvement of type-c individuals in trade unions amount to the difference between the social gains derived form unionization, y(s) c + S, and social gains in case of non unionization, y. Social and private gains are equal if the minimum wage satisfies condition (3). Private and social gains are generally not equal because non union members are paid below their productivity ȳ. In this context, the minimum wage is used to induce the efficient level of union density. The minimum wage has to be strictly positive only if the bargaining power of trade unions, β, is above the threshold value [y(s) ȳ] /y(s), which raises with the level of attitudes towards social cooperation. In that case, private incentives to join unions are too strong in the absence of minimum wage. Therefore, increasing the minimum wage is a means to reduce the costs associated with union wage setting when labor unions do not give rise to enough increase in productivity. When attitudes towards social cooperation improve, the optimal minimum wage level is lower since the rise in productivity induced by trade unions are also higher. Result 2: The equilibrium trade union density increases with the strength of attitudes towards social cooperation. This result is obtained by computing the derivative of union density D defined inequation (1) for the optimal value of the minimum wage given by equation (3). In that situation, one gets ½ F (S + y(s) ȳ) if S< S D = (4) F (βy(s) +S) otherwise. 12 It can easily be checked that second-order conditons are fulfilled. Moreover, c >0 is always satisfied at the optimum. 9

12 h i where S, which satisfies βy( S)+ ȳ y( S) =0, is the threshold value of the strength of attitudes towards social cooperation above which the minimum wage amounts to zero. Welfare and employment The influence of attitudes towards social cooperation on welfare and employment can be summarized by the two following results: Result 3: Aggregate welfare increases with the strength of attitudes towards social cooperation. It can easily be checked that improvements in attitudes towards social cooperation increase welfare by computing the derivative of the objective of the government (equation (2)) with respect to S for the equilibrium values of union density and minimum wage. One gets dw ds = D 1+y 0 (S) > 0. This equation shows that attitudes towards social cooperation improve welfare for two reasons. First, they increase the benefits that individuals get from collective action. Second, they help to improve productivity. Thepositiveeffect of attitudes towards social cooperation on welfare might also affect the employment patterns when one assumes that labor market participation is endogenous. The most natural way to introduce endogenous labor supply in this context is to consider a distribution of indirect utilities outside the labor market, whose cumulative distribution function, denoted by H, is continuous over its support. For the sake of simplicity, we assume that the distribution H is independent of the distribution F over the costs of unionization c. Moreover, it is assumed that individuals who enter into the labor market and get no job (because they refuse the take-it of leave-it offer of the employer when they are not unionized) get zero gains, and then reach an indirect utility level equals to zero. In such a framework, employers offer the minimum wage and the equilibrium values of the union density and the minimum wage are the same as those derived when labor market participation is exogenous. Assuming, for the sake of simplicity, that the subjective costs of union membership are known by individuals only after they have decided to enter into the labor market, one can claim that Result 4: Employment increases with the strength of attitudes towards social cooperation. This results holds because labor supply, equal to employment at the equilibrium, amounts to H(W ), where W denotes the optimum value of the objective of the government which increases with social attitudes S. The dynamics of social attitudes 10

13 So far, the analysis was devoted to the short run equilibrium, in which the set of attitudes towards social cooperation is a predetermined variable. However, attitudes towards social cooperation can evolve over time. Following Akerlof (1980), Corneo (1995, 1997) and Lindbeck et al. (1999) it is assumed that the values responsible for the compliance to social custom are less likely to be passed on from one generation to the next one when disobedience is greater. Accordingly, attitudes towards social cooperation follow the law of motion S t+1 = S t + g(x + D t S t ), (5) where g is a continuous and increasing function that satisfies g(0) = 0; X R is a parameter that represents the potential influence of structural factors other than trade union density on the evolution of social attitudes. Substituting the equilibrium union density in period t, defined in equation (4) into equation (5), one gets the equation ½ g(x + F (St + y(s S t+1 = S t + t ) ȳ) S t ) if S< S g(x + F (βy(s t )+S t ) S t ) otherwise, where S satisfies βy( S)+ h y( S) i ȳ =0. This equation determines the dynamics of social attitudes on the equilibrium path. The dynamic properties hinge on the properties of the functions F and g. Previous assumptions imply that there always exists a steady state equilibrium with S =0, which corresponds to the minimum level of attitudes towards social cooperation, with zero union density and a positive minimum wage equals to βȳ. However, this equilibrium is not necessarily stable. Figure 4 displays a situation in which the equilibrium with the minimum level of attitudes towards social cooperation is stable at low value of X but becomes instable as X takes on higher values. When the equilibrium with S =0is unstable, there exists another stable steady state equilibrium in which both the minimum wage and union density are strictly positive. This representation of the dynamics of social attitudes suggests that countries can exhibit very different dynamics. Some countries, whose value of X is low, can start with strong attitudes towards social cooperation and then face continuous decreases in attitudes towards social cooperation, accompanied by union density drops and minimum wage increases. Other countries, whose value of X is large, can start from weak attitudes towards social cooperation but will eventually converge towards high social cooperation and union density. 3 Empirical results Let us now analyze the empirical relevance of the model by estimating the link between attitudes towards social cooperation and wage-setting institutions in OECD countries since the early 11

14 St S t Figure 4: The dynamics of social attitudes S t when, ȳ =1,y(S) =ȳ+0.1 S, β =0.5,g(x) =x/2, and F is Log Normal with mean and variance equal to one. The thick dotted line corresponds tothecasewherex =0. The thick continuous line corresponds to the case where X =0.3 and the thin continous line is the 45 line. In both cases, the equilibrium minimum wage is strictly positive if S t is smaller than

15 1970s. Our empirical strategy consists first in documenting the wide cross-country differences in attitudes towards social cooperation which can be a potential source of heterogeneity in labor market institutions. Second, we document the empirical relevance of our assumption according to which attitudes towards social cooperation are not instantaneously overdetermined by the economic and institutional features of countries in which people are living. Instead, attitudes towards social cooperation turn out to be partly ingrained in national features with long-lasting effects transmitted across generations even when individuals no longer live in the country of origin. Third, we draw upon this result to show that attitudes towards social cooperation significantly affect contemporaneous national wage-setting institutions and labor market performance by controlling for potential reverse-causality effect. 3.1 Attitudes towards social cooperation and attitudes towards labor market institutions In this section, we document the cross-country heterogeneity in attitudes towards social cooperation. We assess the extent to which this cross-country heterogeneity in consistently correlated withvariationinpreferencestowardsdifferent wage-setting institutions. We then examine the underlying forces which drive this cross-country correlation by stressing the overwhelming role of the national context in shaping individual attitudes Data The cross-country measures of the correlation between attitudes towards social cooperation and attitudes towards labor market institutions come from the World Value Survey database (WVS) and the International Social Survey Programme database (ISSP). These studies sampled the attitudes of publics in all OECD countries at least over the eighties and nineties. The WVS database covers three main waves (1981, 1990, ) while the ISSP study conducted surveys on specific topicsateachwave. The first obvious feature of attitudes towards social cooperation which matters in the realm of collective action and adhesion to organizations is the level of trust in others. In bargaining, interactions between trusting individuals or organizations might lead to efficient outcome, whereas lack of trust might require interventions from an outside person to overcome inefficient equilibrium. To tap this sense of social cooperation among citizens, we use the question on the trustfulness of people provided by the WVS database: Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?. Our indicator trust is the percentage of respondents in each country replying Most people can be trusted instead of One must be careful (after deleting the answers do not know ). 13

16 A second key ingredient of attitudes towards social cooperation which is likely to affect cooperation on the labor market and the political support for minimum wage is the feeling of exploitation or fairness. The fairness item is assessed by the question: Do you think most people try to take advantage of you if they got a chance or would they try to be fair?. This fairness indicator is given by the percentage of respondents who answer People try to be fair rather than People would take advantage if they got a chance. The WVS database reports this question for the wave and for most OECD countries. We fill the missing countries by using the ISSP database of 1998 in which the respondents were asked: How often do you think that people would try to take advantage of you if they got a chance and how often they would try to be fair?. The answers are given on a scale of 1 to 4, which correspond to Try to take advantage almost all the time, Try to take advantage most of the time, Try to be fair most of the time and Try to be fair almost all of the time. To ease the comparison of the results with the WVS database, we group the answers together to represent individuals who tend to believe in fairness. Hence we create a dummy variable which takes on the value 1 if the respondent answers Try to be fair most of the time or Try to be fair almost all of the time, and 0 otherwise. Concerning attitudes towards labor market institutions, we are primarily interested in measuring to what extent people are implied in collective action rather than relying on the government to regulate the labor market and wages in particular. For that purpose, we first assess the levelofconfidence or cooperation in collective action by directly measuring union membership. The unionization item is given by the question To which voluntary organizations or activities do you belong? Labor union. This question is provided for all waves in the WVS and ISSP databases. We construct an indicator of union membership which is equal 1 if the respondent is an active or inactive union member, and 0 otherwise. To assess the degree of preference for wage regulation by the government, we use the question: Here are some things the government might do for the economy. Please show which actions you are in favor of and which you are against of. Government should control wages by law?. This question is provided by two special ISSP databases on the role of government in 1991 and The answers are given on an ordered scale from 1 to 5, corresponding to Strongly agree, In favor of, Neither in favor of nor against, Against, Strongly against. Our measure of state intervention to settle down wages is given by the percentage of respondents who answered Strongly agree or In favor of (after deleting the neither agree nor disagree answers). Our empirical investigation on the cross-country correlation between attitudes towards social cooperation and attitudes towards labor market institutions is based on the working age population and includes 22 OECD countries: Australia, Austria, Belgium, Canada, Czech Republic, 14

17 Denmark, France, Germany, Hungary, Ireland, Italy, Japan, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Spain, Sweden, UK and USA. By grouping the different countries and different waves together, this selection leaves us with a sample of 60,607 working aged individuals in the WVS database and 25,280 working aged individuals in the ISSP database. The summary statistics for the number of observations by country and the individual characteristics are reported in Table 6 and Table 7 in Appendix Cross-country correlations between attitudes towards social cooperation and towards labor market institutions We firstinvestigatetowhatextenttrustandfairness, the two components of attitudes towards social cooperation that we account for, are correlated with the two attitudes towards labor market institutions we are interested in: individual union membership and individual support for wage-setting by law. To get a synthetic measure of attitudes towards social cooperation, we define a single social cooperation value for each nation by taking the means of the individual reply over the two questions about trust and fairness. The decomposition by indicators reports the same cross-country correlation pattern as shown in Appendix. Figure 5 shows the cross-country correlation between the composite indicator of attitudes towards social cooperation and attitudes on labor market regulation. First, the x-axis of Figure 5 shows the fairly high level of cross-country heterogeneity in the mean level of attitudes towards social cooperation. On average, about 70 percent of the population in Norway, Finland, Sweden, Denmark and Netherlands have strong attitudes towards social cooperation. By contrast, Southern European countries are characterized by fairly low social attitudes. They are in general less than one third of the population to think that people can either be trusted or are fair in countries like France, Spain,Italy, not to say Mexico and Turkey. To a large extent, respondents living in Eastern European countries share the same low level of social attitudes as their Southern European counterparts. Eventually, Anglo-Saxon countries such as the United States and United Kingdom, and European Continental countries such as Austria and Germany, stand at an intermediate position. For instance, about 69 percent of German people and 65 percent of American people believe that people would try to be fair instead rather than exploiting others even if they have a chance to do it. Second, the y-axis of Figure 5 shows that this heterogeneity in attitudes towards social cooperation is highly correlated with the same cross-country heterogeneity in attitudes towards labor market institutions. The y-axis on the left panel reports the percentage of respondents who belong to a labor union. This percentage is much higher in Nordic countries, reaching more than 50 percent in Denmark and Sweden, to be compared with less than 10 percent of 15

18 Labor union member (%) Mx Bg Pt ItaFra Sp Uk Germ Cd Usa Aus Aut JpIre Fin Swd Dk R²=0.475 Nth Nw Wage-setting by Law (%) Ita Sp Fra Czr Hg Uk Cd Germ Usa Jp Aus Ire Swd R² =0.338 Nw Social cooperation(% ) Social c ooperation (%) Figure 5: Correlation between social attitudes and labor market attitudes. Source: WVS , ISSP 1996, union membership in Southern countries. Moreover, the correlation is quite high between union membership and attitudes towards social cooperation (R 2 =0.475). The same cross-country heterogeneity holds regarding the mean national level of respondents who consider that wages should be directly set by laws. In general Anglo-Saxon and Scandinavian countries strongly opposed such a legislation whereas Southern countries such as Italy and Spain strongly back this policy. Significantly, opinions in favor of a state regulation of wages is negatively correlated (R 2 =0.338) with our composite indicator of attitudes towards social cooperation. Third, Figure 5 suggests a strong negative correlation between unionization and advocating for wage-setting by government. This correlation seems to reflect the fact that people engaged in a labor union seem to believe more in collective and cooperative action to regulate the labor market and wages rather than relying on the intervention of the government to set wages by law. 16

19 3.1.3 The role of national features Next, we explore the underlying features driving the correlation between attitudes towards social cooperation and attitudes towards labor market institutions. Is such a correlation linked to specific individual characteristics or to national features? We address this issue by running probit estimates on the different indicators for attitudes towards social cooperation and attitudes towards labor market institutions. The role of specific national features is measured by using country-fixed effects. We also control for individual characteristics including gender, age and age squared, the number of years of education, employment status, income category, union membership, political orientation and religious affiliation. Table 1 reports the results of the probit estimates for the attitudes towards social cooperation indicators (trust, fairness). The left-hand side variables of the regressions are dummies equal 1 if the respondent agrees with our main statement of interest regarding social cooperation and labor market attitudes. In all regressions, we take Sweden as the reference country. The latter one has the advantage to provide a benchmark country with one of the highest level of attitudes towards social cooperation and to be represented for all the indicators we are interested in. Strikingly, all the country dummies are jointly highly significant as reported in Table 1. Figure 6 illustrates this result by looking at the marginal coefficients of each country relatively to Sweden. To get a comprehensive indicator on attitudes towards social cooperation, we still use a composite indicator as the means of the marginal effects estimated for the two original indicator trust and fairness. The first striking result is the quantitative importance of the estimated country fixed effects. Figure 6 shows that the fact to live in a Mediterranean countries like France or Italy reduces the level of attitudes towards social cooperation by 24 percent and 26 percent by comparison with an individual sharing the same characteristics but living in Sweden. By contrast, living in another Nordic countries might increase the level of social cooperation by 4 percent in the Netherlands or 8 percent in Norway. Anglo-Saxon and European continental countries lie in between. Living in Germany or the United States decreases the level of attitudes towards social cooperation by 13.8 percent and 14.2 percent relatively to someone living in Sweden. The same opposition pattern holds regarding attitudes on labor market regulation. Living in a Mediterranean country like Spain reduces the probability of belonging to a union by 23 percent relatively to people with the same characteristics in Sweden while it increases the probability to back wage-setting by law by 39 percent relatively to Sweden. The natural outcome of these regressions is that the whole bulk of the cross-country correlation between social cooperation and labor market attitudes is driven by national specific features. Figure 6 shows that the correlation between the estimated country fixed effects is fairly identical 17

20 Table 1: Social attitudes in OECD countries Dependent variable Trust in other=1 People try to be fair=1 Coeff Std Error Coeff Std Error Country dummies Yes *** Male.013 (.014) *** (.055) Age.009 *** (.002).063 *** (.019) Age *** (.000) *** (.000) Education.076 *** (.028).039 *** (.008) Employed Reference Unemployed (.076) (.139) Inactive (.020).107 (.070) Income: mid Reference Low *** (.018) ** (.060) High.110 *** (.018).200 ** (.096) Political orientation: Center Reference Left.125 *** (.017) (.067) Right.015 (.016) (.076) Religious affiliation: No_religion Reference Catholic *** (.023).022 (.077) Protestant.070 ** (.029).084 (.090) Buddhist (.067) (.342) Muslim (.105) Jews.337 *** (.115) Other_religion.018 (.053).086 (.162) Pseudo-R Observations ***:1%, **: 5%, *: 10% WVS 1981, 1990, 2000 ISSP

21 Country effects: Labor union member e-16 Mx Bg Fr a Ita Sp Pt Cd Usa Uk Germ Aut Aus JpIre Fi n Swd Dk Nth Nw R²=0.47 Country effects: Wage-setting by Law Fra Sp Ita Czr Cd Hg Uk Germ Usa Aus Jp Ire Swd Nw R² = e Country ef fects: Social cooperation e Country ef f ects: Social cooperation Figure 6: Country fixed effects in social attitudes and labor market attitudes. Source: WVS , ISSP tothatdisplayedbythemeanreplyinfigure5. Moreover, the variance in the cross-country correlation is almost entirely accounted for by the country fixed effects, the R-squared remaining fairly unchanged when one excludes individual characteristics from the correlation pattern. 3.2 Social cooperation and Labor market outcomes Estimation strategy How do attitudes towards social cooperation relate to wage-setting institutions? And do they explain cross-country differences in labor market performance? Our goal is to answer this question by looking at the issues raised by the estimation of the following linear equation based on the predictions of the theoretical model: I ct = α 0 + α 1 S ct + α 2 X ct + α 3 F c + α 4 T t + ε ct (6) where I ct stands for the wage-setting institutions represented either by union density or by minimum wage legislation in country c at period t, S ct measures country average of attitudes towards social cooperation, X ct denotes a vector of average characteristics of the population and 19

22 of the economy, F c stands for country dummies capturing all other specific featuressuchasthe legal origins or past institutions with long-lasting effects, T t stand for period dummies; ε ct is an error term. The problem with equation (6) is that contemporaneous attitudes towards social cooperation arelikelytobeinfluenced by current institutions. As suggested by the model, individuals living in an environment with a high level of union density might be more prone to cooperate with others.orpeoplelivinginacountrywithveryweakminimumwagelegislationandweaklabor unions might have a greater feeling of exploitation and have low expectations regarding the fairness of other people. We are thus looking for variables that influence social attitudes but which are exogenous as regards current institutions and labor market performance. In other words, we are looking for the inherited part of national social attitudes which is ingrained in individuals independently of the contemporaneous national institutions. In order to find such a variable, we focus on individuals currently living in the United- States, but who differ by the country of origin of their ancestors. We then measure the impact of the country of origin of their forebears on their current attitudes towards social cooperation, controlling for their other individual characteristics and the economic environment. This strategy leads us to estimate the equation where s US t s US t = β US 0 + β US ct c US t + β US 2 x US t + β US 3 Tt US + η US t, (7) stands for the individual attitudes of people living in the US in period t, c US t a dummy variable indicating their ancestor s country of origin, x US t is a vector of individual characteristics, Tt US stands for period dummies, and η US t is an error term. In this context, the variable S ct which shows up in equation (6) and which denotes average attitudes towards social cooperation in country c at time t can be decomposed into two terms. First, a term which accounts for the attitudes inherited from previous generations independently of the current social and economic environment, which corresponds to the coefficient 13 β US ct in equation (7). Second, a residual term, denoted by R ct, which accounts for all other elements which influence social attitudes. Accordingly, one can write Then, using this expression for S ct, we estimate the equation S ct = β US ct + R ct. (8) I ct = γ 0 + γ 1 β US ct + γ 2 X ct + γ 3 F c + γ 4 T t + ν ct, (9) 13 More formally, β US ct is equal to the difference E s US t c t 6= rc, x US t,tt US E s US t c t = rc, x US t,tt US where E denotes the expectation operator and rc stands for the country of origin chosen as the reference country. is 20

23 where the coefficient γ 1 measures the impact of the attitudes inherited from previous generations on current institutions. The implementation of this strategy is described in the next section. But before turning to this point, it is worthwhile to stress what we can expect from such an identification strategy. Since we are able to identify the attitudes inherited from previous generations, we no longer have to worry about a potential endogeneity bias stemming from the fact that current cultural attitudes are endogenous to contemporaneous institutions and employment patterns. But have we identified a causal effect going from inherited current attitudes to current institutions? The answer might not be a qualified yes. Attitudes towards social cooperation of earlier generations who immigrated to the United States in the wake of the 20th century could still capture an omitted historical variable which would explain both attitudes of earlier generations and contemporaneous labor market institutions. Even by focusing on Americans whose forebears immigrated in the late 19th century at a time when unions and minimum wage legislation were not formally implemented, it might still be the case that earlier laws on guilds and cooperation, or any other country specific feature, could have influenced both forbears social capital and contemporaneous national wage-setting institutions. The introduction of country fixed effects in equation (9) is precisely meant to control for such omitted variables Inherited attitudes Data We estimate the national cultural traits of earlier generations by using the General social Survey. This database consists of an individual surveys on the United States over the period The first attraction of the GSS data is that they contain questions and ordered scale answers on social attitudes which are exactly identical to those asked in the WVS and ISSP data. We are thus able to construct two similar indicators on trust and fairness in the United States. The question on the trustfulness of people reads: Generally speaking, would you say that most people can be trusted or that you can t be too careful in dealing with people?. The answers are given on a scale from 1 to 3, which correspond to Most people can be trusted, Can t be too careful, Depends. In lines with our cross-country comparison, the trust indicator trust_usa is a dummy variable equal to one if the respondent answers that people can be trusted and 0 if she answers that one should be careful. Regarding fairness, the question reads: Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?. The answer also ranges from 1 to 3 corresponding to Would take advantage of you, Would try to be fair and Depends. Our indicator fairness_usa is measured as the 21

24 percentage of respondents who answer that People would try to be fair instead of Would take advantage of you. As the prior analysis, we focus on a comprehensive social attitudes_usa indicator measured as the average of trust_usa and fairness_usa indicators. To sort out the role of earlier generations on cultural traits, we use informations on the country of origins of the respondent s forbears. The GSS ethnic variable reads as follows: From what countries or part of the world did your ancestors come from?. Respondents are also asked if they are born in the United States and how many of their parents and grand-parents were born in the country. The question on parents birthplace is scaled 0 if both parents are born in the USA, 1 if only the mother is born in the US, and 2 if only the respondent s father is born in the country. The answer on grand-parents birthplace is scaled from 0 to 4 indicating the number of grandparents born in the US. This information allows us to a certain extent to control for the transmission durability of cultural features by identifying the wave of immigration of the forbears. Yet it is important to stress that it is still an approximate measure of ethnic heritage since the GSS asks respondents to name only a single country from which their forbears immigrated. But many Americans of course have ancestral ties to more than one country. Yet another key attraction of the GSS data is the sample size of the ethnic variable. Most European ethnies are represented including 6108 observations for respondents with German origins, 5754 observations for Anglo-Saxon origins, 1836 observations for Italian origins, 722 observations for French origins and 669 observations for Norvegian origins. Additionally, the GSS database makes it possible to control for the main other socioeconomic characteristics as those defined in the WVS and ISSP data including age, gender, employment status, religious affiliation, political orientation. Significantly, the GSS data also asks respondents the level of education of their parents. This information might be crucial since potential correlation between social attitudes and ethnic heritage might transit through parents characteristics such as human capital rather than culture per se. In lines with our previous cross-country analysis, our main sample consists of working age people who are years old. To control for the cultural part of social attitudes inherited from parents independently from contemporaneous national institutions, we only select people who are born in the United States. Since respondents were asked about their birthplace only since 1977, our sample covers the period We only focus on respondents whose ancestors came from OECD countries including Austria, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Japan, Mexico, Netherlands, Norway, Poland, Portugal, Spain, Sweden, and the United Kingdom. Our main sample combining informations on the country of ancestry and the level of social capital consists of 20,354 observations. Detailed descriptions of the countries of ancestry and the summary statistics for individual characteristics 22